I've made a YouTube video with clips from a single conversation I had this morning on the repeater. The audio was recorded on my dashcam.
As can be heard in the video, when I start the conversation, the audio delay is nil, or very little when unkeying. However, by the time I'm finished (I think the conversation lasted about 20 minutes) the audio delay is out to 4 or 5 words!
We were both using the same repeater. We have a repeater linked in from another location. I've noticed when talking to someone via that method, the audio delay issue does not build up so quickly.
Hopefully this explains a bit better the issue I'm having.